xml_index

#!/usr/bin/python
import xml.sax as sax
import xml.sax.handler as handler
from optparse import OptionParser
import sys
import pprint
import re
import codecs
import os

class IndexHandler( handler.ContentHandler ):
    def __init__(self,verbose = False):
        handler.ContentHandler.__init__(self)
        self.verbose = verbose
        self.summary = {}
        self.maxlen = {}
        self.sumlen = {}
        self.common = {}
        
    def startElement(self,name,attrs):
        if name == "Metadata":    
            mrtype = attrs['MYRADIOTYPE']
            for k in attrs.keys():
                key = mrtype+"."+k
                if key in self.summary:
                    self.summary[key] += 1          
                    self.sumlen[key] += len(attrs[k])   
                    if len(attrs[k])/10 in self.common[key]:
                        self.common[key][len(attrs[k])/10] += 1
                    else:
                        self.common[key][len(attrs[k])/10] = 1
                    if len(attrs[k]) > self.maxlen[key]:
                        self.maxlen[key] = len(attrs[k])
                else:
                    self.summary[key] = 1
                    self.maxlen[key] = len(attrs[k])
                    self.sumlen[key] = len(attrs[k])
                    self.common[key] = {}
                    self.common[key][len(attrs[k])/10] = 1
        
    def characters(self, content):
        pass
        
    def endElement(self,name):
        pass
            

if __name__ == '__main__':
    parser = OptionParser(usage="usage: %prog [options]", description="index music information from xml files")
    parser.add_option("-v","--verbose", dest="verbose", action="store_true", default=False, help="turn on debug output")
    
    (options,args) = parser.parse_args()
    
    if len(args) < 1:
        sys.exit("Must specify at least one path with xml files")

    errlog = open("xml_error.log","w")
    handler = IndexHandler(options.verbose)
    for (dirpath, dirnames, filenames) in os.walk(args[0]):
        for f in filenames:
            if re.match("^.*\.xml$",f):
                if options.verbose:
                    print "Process",dirpath,f
                try:
                    sax.parse(open(os.path.join(dirpath,f)), handler)
                except Exception,e:
                    print >>errlog,"File",os.path.join(dirpath,f),"Error",str(e)
    errlog.close()
    
    outcsv = codecs.open("xml_index.csv","w","utf-8","replace")
    for (key,count) in handler.summary.items():
        print >>outcsv,'"%s",%d,%d,%f,"%s"'%(key,count,handler.maxlen[key],float(handler.sumlen[key])/count,str(handler.common[key]))
    outcsv.close()